Claude Code Common Commands

In Claude Code, simply type / to see all currently available commands.
Continue typing / followed by letters to quickly filter commands.

The official documentation also specifically notes: not all commands are visible to every user.
Some commands are affected by the platform, plan, environment, or terminal capabilities.

If you are just starting to use Claude Code, you don't need to memorize dozens of commands right away.
Mastering these 15 is enough to get you started:

Command

Purpose

Tip

/help

View help and available commands

Look here first if you forget a command

/clear

Clear current conversation history

Very useful when the discussion goes off track

/config

Open settings interface

Change models, themes, and output styles

/status

Check current status

View version, model, account, and connection

/model

Switch models

Use when you want to change the capability tier

/usage

Check plan usage and rate limit status

When you care about quotas and rate limits

/cost

View token and cost statistics

When you want to control costs

/permissions

View or update permissions

Control whether tools are allowed to execute

/plan

Enter plan mode

Plan big tasks first before executing

/diff

View differences of current changes

Very useful after modifying code

/memory

Manage CLAUDE.md and memory

All long-term project constraints are here

/context

View context usage

Troubleshoot context issues after long conversations

/resume

Resume previous session

Continue from where you left off last time

/compact

Compact conversation context

Very useful after long conversations

/rewind

Rewind to a previous checkpoint

Very useful when things get broken

/hooks

View tool event hooks configuration

When you want to implement automated enhancements

This category of commands mainly solves the problem of "how to manage the current session".

Clear conversation history and free up context.
The official documentation also mentions it has two aliases:

• /reset
• /new

When is it suitable to use:

• The current conversation has already become cluttered
• You want to switch tasks and start fresh
• The context has been polluted by old questions

Resume historical sessions.
You can resume via session ID or name, or open a selector to continue.

Rename the current session.
If you don't specify a name, Claude Code can automatically generate one based on the conversation history.

Exit the CLI.
The alias is /quit.

This category of commands is closest to the actual development workflow.

The official documentation makes it very clear: it can directly enter plan mode from a prompt.
You can also optionally include a description, for example:

/plan fix the auth bug

This is perfect for:

• Changes that span multiple files
• You want to see the plan first, without having Claude modify the code immediately
• You want the task to be broken down clearly first

Open the interactive diff viewer.
You can use it to view:

• Current uncommitted Git changes
• Differences generated by Claude in each round

This is much more reliable than simply asking, "What did you change?"

List and manage background tasks.
If you have Claude running a task that takes a long time, this command will be very useful.

Fetch and display GitHub Pull Request comments.
The official documentation notes that it depends on the gh CLI.

This category of commands resolves how Claude Code is currently running.

Open the settings interface.
The official documentation mentions that it can adjust:

• Themes
• Models
• Output styles
• Other preference settings

The alias is /settings.

Open the status interface to view:

• Version
• Model
• Account
• Connectivity

The official documentation specifically highlights: it also works while Claude is responding, without needing to wait for the current response to finish.

This is the command you specifically requested to add earlier.
The definition given in the official documentation is very straightforward:

Display plan usage limits and rate limit status

It is suitable for use at times like these:

• You suspect you are about to hit your plan limits
• You want to check if you are currently being throttled by rate limits
• You have been using it heavily for a while and want to check your quota status

If you often use Claude Code heavily, keeping this command handy is recommended.

The official documentation says: Display token usage statistics.
If you are concerned about costs, this command goes hand-in-hand with /usage:

• /usage focuses more on plan limits and rate limits
• /cost focuses more on token consumption and session costs

Select or switch models.
For supported models, you can also further adjust the thinking effort level.

Set the model effort level.
This command is important for advanced users, as it directly affects Claude's depth of thinking and response cost.

View or update permissions.
If you find Claude always getting blocked on certain tool calls, this command is crucial.

Switch color themes.
The official documentation mentions that it supports:

• Light and dark variants
• Colorblind-friendly themes
• ANSI themes using terminal colors

This isn't a core productivity command, but if you use Claude Code for long periods, fine-tuning the interface experience is well worth it.

Toggle between Vim and regular editing modes.
If you are already accustomed to Vim-like operations in your terminal, this command will be quite practical.

The official documentation states very clearly:
It initializes the project using the CLAUDE.md guide.

This is generally suitable for:

• Introducing Claude Code to a project for the first time
• Systematically establishing project instructions and work constraints

This command is extremely important.
It allows you to:

• Edit CLAUDE.md
• Enable or disable auto-memory
• View auto-memory entries

You can think of it as Claude Code's "project memory management portal."

The official documentation describes it as:

Visualize current context usage as a colored grid. Display optimization recommendations for context-intensive tools, memory bloat, and capacity warnings

In other words, instead of simply telling you "how many tokens are left," it helps you clearly see:

• Whether the current context is already getting very heavy
• Which tools or historical contents consume the most context
• Whether memory bloat is occurring
• Whether you are close to the capacity boundary

This is extremely useful in long conversations.

• The conversation has gone on for many rounds
• You feel like Claude is starting to "forget previous context"
• Response quality declines, but you don't know if the context is too full
• You want to determine whether to execute /compact

You can think of it as Claude Code's "context health check command."

The official documentation says: Compact the conversation, with the option to pass in focus instructions.
This is a very practical command in long sessions.

For example, you can understand it like this:

• The conversation is too long
• The context is starting to bloat
• You want Claude to keep the key points and discard redundancies

At this point, you can use /compact, optionally appending a focus for compression if needed.

/compact 只保留和支付模块改动有关的上下文

Manage MCP server connections and OAuth authentication.

Manage Claude Code plugins.

Manage IDE integrations and view status.

Enable remote control of the current session from claude.ai.
The alias is /rc.

If you didn't realize before that Claude Code has remote bridging capabilities, this command is one of the most direct entry points.

The official documentation describes it as: View tool event hook configurations.
If you want to implement more advanced automation, such as:

• Triggering scripts before or after tool execution
• Adding automatic checks
• Integrating with your own engineering workflow

This command is a very important entry point.

Copy the most recent assistant response.
If you pass a number N, you can copy the Nth most recent response.

Export the current conversation as plain text.
Suitable for archiving, reviewing, and preserving work logs.

Create a branch for the current conversation.
The alias is /fork.

Diagnose the installation and setup of Claude Code.
If you suspect an issue with your environment, this is a high-priority troubleshooting entry point.

The official documentation states that it can:

• Rewind the conversation to a previous point
• Rewind the code to a previous state
• Or summarize from a selected message

The alias is /checkpoint.

This command is particularly suitable for:

• When a certain round of changes doesn't turn out right
• When you want to return to an earlier safe point
• When you need to re-implement a certain part of the path

If you have already passed the beginner stage, these commands are also well worth gradually mastering:

Command

Purpose

/copy [N]

Copy the most recent or Nth most recent response

/export [filename]

Export current conversation

/branch [name]

Create a branch for the current conversation

/doctor

Diagnose installation and configuration issues

/context

Visualize context usage and sources of bloat

/theme

Adjust color themes

/vim

Toggle Vim mode

/tasks

Manage background tasks

/pr-comments [PR]

Fetch GitHub PR comments

/ide

View and manage IDE integrations

/remote-control

Enable remote control

You can absolutely use them like this in your daily workflow:

/plan
先规划这个需求，不要立刻改代码。

确认后开始修改。

/diff
看一下本轮改动。

/usage
确认一下当前使用限制状态。

/memory
把这次沉淀下来的项目规则写进 CLAUDE.md。

There are many official commands, but you don't need to know them all at the beginning.
A more logical learning path is:

1. Master session control first
2. Then master configuration, status, /usage, and /cost
3. Then master /plan, /diff, /permissions, and /compact
4. Then master /memory and /rewind
5. Finally, move on to extension capabilities like MCP, plugin, hooks, and remote-control

If you can't remember them, just go back to the very first sentence:

Type / in Claude Code to see which commands are currently available, then choose based on your scenario.